Multi-Capture High Dynamic Range Compressive Imaging

We propose a novel approach to reconstruct High Dynamic Range images from few compressive measurements. The reconstruction algorithm directly merges the information of multi-capture bayerized images. It simultaneously performs demosaicing and naive predefined tone-mapping. Two different color spaces are taken into account at the reconstruction stage to add multiple constraints on the signal. The proposed method provides a new way to compensate nonlinearities due to saturation issues in the case of multi-capture. This technique requires some prior knowledge of the pixel behavior at the sensor level. Yet, it could already be applied using existing compressive measurements performed by actual Compressive Sensing image sensors such as [1]. Future compressive sensing imagers would be able to efficiently acquire multi-capture high dynamic range compressed images.